UnfairFace
收藏arXiv2024-04-29 更新2024-06-21 收录
下载链接：
https://github.com/MikeLasz/Benchmarking-Fairness-ImageUpsampling
下载链接
链接失效反馈官方服务：
资源简介：
UnfairFace数据集是FairFace的一个子集，专门设计来复制常见大规模人脸数据集的种族分布。该数据集包含约20000个样本，其中大多数被标记为'White'，用于评估和基准测试图像上采样方法的公平性。通过这个数据集，研究者们能够深入探讨算法在处理不同种族群体时的表现差异，特别是在少数群体样本的重建上。
UnfairFace dataset is a subset of FairFace, specifically designed to replicate the racial distribution of prevalent large-scale face datasets. It contains approximately 20,000 samples, the majority of which are labeled 'White', and is utilized for evaluating and benchmarking the fairness of image upsampling methods. Through this dataset, researchers can thoroughly investigate the performance differences of algorithms when processing different racial groups, especially regarding the reconstruction of samples from minority groups.
提供机构：
鲁尔大学波鸿分校计算机科学系
创建时间：
2024-01-25
